Parasitic spikes at smps DC output

Hi, so here's my problem , I've built an working smps , delivers loads etc fine but since i'm using for audio it's giving me this problematic output. My output is 90-0-90 VDC.I have two secondaries , two separate transformers each having a bridge rectifier and filter caps , also snubbers across the filter cap bank traces and idle voltage limit resistors across the 90-90 outputs.
both transformers are driven from a single half bridge consisting of two IGBT's driven by IR2110 which gets it's frequency and waveform from SG3525.
both transformer output literally have the same problem , when i probe with my scope into each transformers rectified output no matter whether between 90 to gnd or vice versa I get the same sinusoidal spikes.
I will have a good cam tomorrow I could try to take some screen shots.

I tend to to think that these spikes I see at output are the turn on and off moments of the half bridge switches , somehow the current entering the primary causes a spike.
or that one of my transformers maybe has an airgap in the core since it was taken from an older philips tv power supply but it's not the typical hv ferrite transformer for the high voltage CRT voltage.
Well I will do some more tesing tomorrow without that particular transformer and see what happens.

Meanwhile maybe you guys have some advice.

I cant see any feedback in your circuit.
Also, I am sure you know the 1u blocker cap Is only for voltage mode. (have you set the sg3525 up for voltage mode?)
The resonance frequency of the 1uF and the transformer leakage must be well lower than the switching frequency.

- - - Updated - - -

I believe you should better use say a 10k GS resistors.
(gate to source)
Also, a few ohms res in series with the bootstrap diode.

- - - Updated - - -

isn't 22uF very big for a bootstrap cap?

- - - Updated - - -

also, I would use gate to source zeners.

i dont know what your power level is but i reckon youd be better off with a single full bridge than two halfers. Two halfers is harder to lay out and this could be related to your problem.

- - - Updated - - -

I notice you have no feedback. I have sometimes seen LLC converters with no feedback other than rough overvoltage protection. This is sometimes possible with an LLC, because if kept at the resonant frequency, and with a constant vin, then it acts like a high frequency transformer and just makes the output voltage equal to VIN * NS/NP, over a wide load range.

With a half bridge converter, it is not really possible to do that, because the secondary voltage is above the output voltage, and you will tend to go up in voltage at lighter loads.

So , would you consider the LLC?…..there are LLC’s on the web for guitar amp power supplies……..i think by connexelectronic

**broken link removed**

treez said:

The resonance frequency of the 1uF and the transformer leakage must be well lower than the switching frequency.

Click to expand...

Does anyone have practical advice on how much lower the resonant frequency of the blocking cap and leakage inductance should be?

preferably less than three times the switching frequency

Well I have been sick for these past two days but I have made some progress , I was rather suspicious of one of my two transformer cores , at startup it always jumped in secondary voltage , I mean the other transformer rises normally until it reaches its max idle output, because i've set a pretty slow rise at startup for the square wave.the other transformer always had this inductive spike at startup going up to about 2...3x it's max idle voltage.it also ran hotter under heavy loads.
Now I took the transformer off first electrically disconnected it and guess what now at my other output the noise level went from about 2-3v in amplitude to 0.2 volts in amplitude and is inaudible through the speakers now.

I took out the transformer and dismantled it's core leaving just the bobbin , if i'm not mistaked it's not ferrite but this other material used for smps transformers that run on lower frequencies , this traffo comes from an older philips tv , i forgot the material but it wasn't ferrite it was something else, it also has this weird electronics smell when broken apart.

P.S. that blocking cap is 2.2uF instead of 1

Well treez I took your advice and put 10k G-S resistors and guess what , as I plugged in the thing I didin't even got the chance to see what the scope says about the ringing issue at the gates and also at output , the switching IGBT's went short circuit and the main fuse box in my home also went down so I had to go there and pull back the breakers.

I think the IR2110 is already made with gate to source charges in mind etc and no extra things are necessary since these resistors apparently took away too much juice and probably left the gates either floating for a bit or didint open or close fully and hence resulted in destruction since before I put them in everything worked fine except the noise thing.
luckily I have one extra pair of IGBT's and will replace them and then will replace the transformer with some proper ETD59 , 5200nH rated core and see maybe that helps.

10k g-s would have drawn no more than 1mA out of the bootstrap cap so it couldnt have been that.....because each fet is 50% duty max so its only 0,5mA in reality.....whatever your problem was, it wasnt the 10k g-s resistors.

- - - Updated - - -

A typical value for a bootstrap cap at say 100khz would be say 220nF......and 1mA coming out of 220nF drops just 45mV in 10 microseconds..............i = c.dv/dt
So again, the 10k g-s resistor was not your problem.

- - - Updated - - -

also, 22uF is very big for a bootstrap cap, what is your f(SW)?

There seem to be quite a few different problems to work through here....

The reason for connecting 10K gate source resistors is to tie down the gates at initial power up. Sometimes (?) sudden application of very high voltage to the IGBT collectors can cause the IGBTs to turn on due to gate collector capacitance if there is nothing to tie the gates down.
However the IR2110 has a very effective undervoltage shutdown so it effectively protects the gates at power up, but fitting 10K resistors is still a good design habit to get into.

The lack of voltage feedback has already been commented on.

The positive and negative outputs from both bridge rectifiers absolutely MUST have effective chokes fitted. These are not a luxury, but a very important part of how the circuit functions.

Bootstrap capacitor value is not that critical, although 22uF is very large its not going to create a problem. It does need to be a tantalum not an aluminium electrolytic.

"needs to be a tantalum" why? Al electro will be fine for boot strap cap, and will not fail like tant's are prone too...

- - - Updated - - -

its most likely your scope is picking up common mode noise (possibly magnetic induced noise as well), short your scope probe and connect it to oVDC of the o/p, noise still there? drape the scope lead near the Tx, noise go up and down?

there's your answer...

- - - Updated - - -

p.s. put your snubbers across the IGBT's, not the Tx, same for the o/p diodes...!